[ntp:questions] Linux 2.6 + ntpd + noapic + insanity.

Omri Schwarz ocscwar at h-after-ocsc.mit.edu
Fri Jan 27 15:51:51 UTC 2006


Hi, everyone.

I have a recently acquired Opteron machine
with Nvidia components that on the more recent
kernels decided to let its system clock drift insanely.

Worse than that, it is just not adjusting to the
ntpd packets it is receiving, so the drift continues
no matter how I try to configure ntpd. (RPM
ntp-4.2.0.a.20040617-8 in Fedora.)

ntp.conf is as follows:

server time.mit.edu prefer
broadcastclient
disable auth
restrict default nomodify lowpriotrap

no special arguments added to the ntpd command line.

Starting up ntpd gives me "kernel time sync status 40"
(which is 0x0040  /* clock unsynchronized (rw) , right?)

ntpq -p says:

# ntpq -p
      remote           refid      st t when poll reach   delay   offset 
jitter
==============================================================================
  KERBEROS.MIT.ED .INIT.          16 u    -  256    0    0.000    0.000 
4000.00
  KERBEROS.MIT.ED 18.7.21.64       2 u   52   64  377    0.001  -160379 
43665.4
  SPACE-A.MIT.EDU 132.163.4.103    2 -   50   64  335    0.280  -162516 
83814.9
  space.mit.edu   18.75.2.100      3 -   29   64  356    0.001  -163226 
85007.1
  newline.mit.edu 192.5.41.40      2 -   58   64  166    0.530  -162271 
85321.4
  UNDERSCORE.MIT. 18.145.0.30      2 -    -   64  335    0.001  -163768 
87353.0

Which IIUC correctly means ntpd is discarding all the data it receives
from the net.

SO beyond putting "rdate -s time.mit.edu" in a crontab, with all the
absurdities entailed in that step, what else am I supposed to do?

Much obliged for your help,



-- 
Omri Schwarz --- ocscwar at mit.edu ('h' before war) 
Timeless wisdom of biomedical engineering: "Noise is principally
due to the presence of the patient." -- R.F. Farr




More information about the questions mailing list